With the same configuration, the training results of the model vary significantly.

[guomengqing]

Hello, I am using a model from mmsegmentation to train on a custom dataset. With the same configuration, the only change I made was the path for saving the results. However, the testing results vary significantly, with the IoU difference being as large as 10%. I have already fixed the random seed and used deterministic algorithms. What could be the reason for this? Is there any way to make the results more stable or reproducible?